What is relevance logic? (Q392272)

The aim of this paper is to define what relevance logic is, since the usual definitions leaning on the notions of the ``use of each premise in the proof'' or ``the variable-sharing property'' (VSP) are not clear, in the author's opinion. The definitions proposed are based on an abstract concept of logic built upon the notion of consequence relation. A consequence relation \(\vdash \) is a \textit{Tarskian consequence relation} (tcr) if it is reflexive, monotone and transitive. A (propositional) logic is a pair \textbf{L} = \((\mathcal{L},\vdash )\) where \(\mathcal{L}\) is a propositional language and \(\vdash \) is a tcr which is structural and non-trivial. The notions of \textit{relevant entailment relation} and \textit{relevant s-entailment relation} are special consequence relations central to the paper. Essentially, the idea is to replace monotonicity with SRC and to require that each one of the said relations satisfy \textit{weak consistency} (\(\nvdash p\) -- \(p\) is a propositional variable) and RP (SRC and RP are defined below). The latter relation is between sets of formulas and formulas; the former, between multisets of formulas and formulas. The main properties of these relations used in the paper are the following:{\parindent=0.7cm\begin{itemize}\item[(1)] the basic relevance criterion (BRC): \(\Gamma \vdash \varphi \) whenever \(\Gamma ,\Gamma ^{\prime }\vdash \varphi \) and \(\Gamma ^{\prime }\) has no propositional variables in common with \(\Gamma ,\varphi \); \item[(2)] the strong relevance condition (SRC): if \(\Gamma \) is not empty and \(\Gamma ,\Gamma ^{\prime }\vdash \varphi \), then \(\Gamma ^{\prime }\) and \(\Gamma ,\varphi \) share some propositional variable; \item[(3)] the (usual) replacement property (RP); \item[(4)] the relevant deduction property (RDP): \(\Gamma ,\varphi \vdash \psi \) iff \(\Gamma \vdash \psi \) or \(\Gamma \vdash \varphi \rightarrow \psi \). \end{itemize}} Next, the notions of weak relevant implication and relevant implication are defined. Let \textbf{L} = \((\mathcal{L},\vdash )\) be a logic: {\parindent=0.7cm\begin{itemize}\item[(1)] \(\rightarrow \) is an \textit{implication} for \textbf{L} iff \(\Gamma ,\varphi \vdash _{\text{L}}\psi \) iff \(\Gamma \vdash _{\text{L}}\varphi \rightarrow \psi \); \item[(2)] The relation \(\vdash _{\text{L}}^{\rightarrow }\) induced by \(\rightarrow \) is defined as follows: \(\varphi _{1},\dots,\varphi _{n}\vdash _{\text{L}}^{\rightarrow }\psi \) iff \(\vdash _{\text{L}}\varphi _{1}\rightarrow (\dots\rightarrow (\varphi _{n}\rightarrow \psi )\dots)\). \end{itemize}} We have: \(\rightarrow \) is a \textit{weak relevant implication} of \textbf{L} if \textbf{L} contains \textbf{LL}\(_{\rightarrow }\) and \(\vdash _{\text{L}}^{\rightarrow }\) satisfies RP and SRC. And \(\rightarrow \) is \textit{relevant implication} of \textbf{L} if it is a weak relevant implication of \textbf{L} which satisfies the RDP (\textbf{LL}\(_{\rightarrow }\) is the implicational fragment of linear logic). The minimal logic with a weak relevant implication is \textbf{LL}\(_{\rightarrow }\); the minimal logic with a relevant implication is \textbf{R}\(_{\rightarrow }\) (the implicational fragment of relevance logic \textbf{R} (cf. [\textit{A. R. Anderson} and \textit{N. D. Belnap jun.}, Entailment. The logic of relevance and necessity. Vol. I. Princeton, N. J.: Princeton University Press (1975; Zbl 0323.02030)]). The minimal logics with a weak relevant implication (relevant implication) and an associated negation (internal negation) are the result of adding the axioms \(\varphi \rightarrow \sim \psi \rightarrow \centerdot \psi \rightarrow \sim \varphi \) and \(\sim \sim \varphi \rightarrow \varphi \) to \textbf{LL}\(_{\rightarrow }\) and \textbf{R}\(_{\rightarrow }\). That is, the logics \textbf{LL}\(_{\rightarrow }^{\sim }\) and \textbf{R}\(_{\rightarrow }^{\sim }\). (It is proved that in the case of relevant implication (not in that of weak relevant implication) \(\rightarrow \) and \(\sim \) are unique.) We can now define the concepts of relevance logic and strong relevance logic. A \textit{strong relevance logic} is a finitary logic with a relevant implication having an associated negation. The minimal strong relevant logic is \textbf{R}\(_{\rightarrow }^{\sim }\). On the other hand, a finitary logic \textbf{L} is a \textit{relevance logic} iff {\parindent=0.7cm\begin{itemize}\item[(1)] \textbf{L} contains \textbf{R}\(_{\rightarrow }^{\sim }\); \item[(2)] \textbf{L} is \(\sim \)-consistent; \item[(3)] \(\rightarrow \) has in \textbf{L} the VSP and the RP; \item[(4)] \textbf{L} satisfies the BRC. \end{itemize}} Now, notice that famous strong relevant logics such as \textbf{E} or \textbf{T} [Anderson and Belnap jun., loc. cit.] or useful weak relevant logics such as \textbf{DJ} [\textit{R. Brady}, Universal logic. Stanford, CA: CSLI Publications (2006; Zbl 1234.03010)] are not even ``relevance logics'' , according to these definitions, while the logic \textbf{R} [Anderson and Belnap jun., loc. cit.] is a ``relevance logic'' , but not a ``strong relevance logic''. A ``strong relevance logic'' mentioned in the paper is the logic \textbf{RMI}\(_{\rightarrow }^{\sim }\) defined by the author.